introrgb rev. 3 2004-07-20 1 introduction to rgb image composites [email protected]...

IntroRGB Rev. 3 2004-07-20 1

Introduction to RGB image composites

Introduction to RGB image composites

Author: HansPeter Roesli, MeteoSwiss Locarno

[email protected]

Contributors: Jochen Kerkmann (EUM)Daniel Rosenfeld (HUJ),Marianne König (EUM)NWC SAF

IntroRGB Rev. 3 2004-07-20 2

Four processing and rendering methods:

1. Images of individual channels, using a simple grey wedge or a LUT* for pseudo colours (typical for MFG channels);

2. Differences/ratios of 2 channels, using a simple grey wedge or a LUT for pseudo colours (e.g. fog, ice/snow or vegetation);

3. Quantitative image products using multi-spectral algorithms (e.g. SAFNWC/MSG software package) and a discrete LUT;

4. RGB composites by attributing 2 to 3 channels or channel combinations to individual colour (RGB) beams classification by addition of RGB colour intensities

Basics of displaying MSG/SEVIRI imagesBasics of displaying MSG/SEVIRI images

* see next slide

IntroRGB Rev. 3 2004-07-20 3

LUT – lookup tableLUT – lookup table

Table allowing a display system to map pixel values into colours or grey scale values with a convenient range of brightness and contrast.

E.g. a narrow range of input pixel intensities may be mapped onto the available range of output intensities.

Rather than using the pixel values directly, the value is instead used as an address into a lookup table where the content of the table at that address defines the output colour or grey-scale value.

Typically, lookup tables addresses have 8 bits, allowing 256 separate input entries, and 8 bits for the output values*.

For colour mapping, three separate colour tables are configured for red, blue and green and arranged such that any of the input bits may control any of the output bits for each colour.

* table addresses may have larger ranges (e.g. 10 bits for MSG/SEVIRI), maximum range of output values is imposed by display hardware (8 bits most common)

IntroRGB Rev. 3 2004-07-20 4

Simple display of individual SEVIRI channels

4 solar (on black), 1 solar + IR (on grey), 6 IR (on white)

Simple display of individual SEVIRI channels

4 solar (on black), 1 solar + IR (on grey), 6 IR (on white)

Adequate for viewing information of 3 MFG channels; Not very practical for 12 MSG/SEVIRI channels;More on SEVIRI channels in 00_rgb_part01.ppt

IntroRGB Rev. 3 2004-07-20 5

Rendering of individual SEVIRI channels - solarProper choice of grey wedge


Solar channels rendered as in black & white photography (channel 03 with particular response from ice/snow) physical rendering using lighter shades for higher reflectivity and darker shades for lower reflectivity.

IntroRGB Rev. 3 2004-07-20 6



solar: reflectivity(P mode onlysee next slide)

high

low

clouds

land / sea

IntroRGB Rev. 3 2004-07-20 7

Rendering of individual SEVIRI channels - IR

Proper choice of grey wedge


Proper choice of grey wedgeIR channels rendered either in P or S mode: P mode - grey shades follow intensity of IR

emission: physical rendering with lighter shades for stronger IR emission and darker shades for weaker IR emission;

S mode - inverted P mode (alternatively also annotated with letter “i” for “inverted”) : traditional rendering, compares better to images from solar channels, i.e. clouds appear in light instead of dark shades.

Note: some IR channels have no direct image application but are useful when combined with other channels or used to derive products, e.g. channels 7, 10 and 11.

IntroRGB Rev. 3 2004-07-20 8




Proper choice of grey wedgeIR: emission / brightness temperature

P mode

strong / warm

weak / cold

clouds / more absorption

land / sea / less absorption

IntroRGB Rev. 3 2004-07-20 9




Proper choice of grey wedgeIR: emission / brightness temperature

S or i mode

strong / warm

weak / cold

clouds / more absorption

land / sea / less absorption

IntroRGB Rev. 3 2004-07-20 10

Differences/ratios of 2 channels

Differences/ratios of 2 channels

Simply displaying a larger set of single channels for comparison is neither efficient in mining useful information nor particularly focussed on phenomena of interest;

Displaying specific channel differences or ratios, a simple operation though, improves the situation awareness by enhancing particular phenomenon of interest (e.g. fog or ice clouds) in a particular situation;

Grey-scale rendering (small values in dark or light shades – large values in light or dark shades) is not standardised; mode may be inherited from similar products based on data of other imagers (e.g. AVHRR or MODIS).

IntroRGB Rev. 3 2004-07-20 11

Differences of 2 channels – using b/w LUT

Differences of 2 channels – using b/w LUT

night - dark day - bright

04 – 09fog

03 – 01ice clouds

day (only) - dark

IntroRGB Rev. 3 2004-07-20 12

Differences of 2 channels – using colour LUT

Differences of 2 channels – using colour LUT

Desert(cloud-free)

Ocean(cloud-free)

Thick IceClouds

Thin IceClouds

Desert Dustor Low Clouds

04 – 09ice / low cloudsdesert dust

IntroRGB Rev. 3 2004-07-20 13

Some recommended differences

Some recommended differences

Clouds 03-01 04-09 05-06 05-09 06-09

Thin cirrus 07-09 04-09 10-09

Fog 09-04 09-07

Snow 03-01

Volcanic ash (SO2) 06-11

Dust 04-09 07-09 10-09

Vegetation 02-01

Fire 04-09

Smoke 03-01More on recommended

differences and their interpretation in other chapters of the Guide

IntroRGB Rev. 3 2004-07-20 14

Quantitative image products using multi-spectral

algorithms


algorithms Quantitative algorithms (thresholding or pattern

recognition techniques) extract specific features from multi-spectral images and code them into a single-channel image quantitative image products;

Using discrete LUTs quantitative images are easy to read due to relation between identified features and colour values, but may have some drawbacks: Feature boundaries appear very artificial (e.g. checker

board due to use of ancillary data of different spatial scale);

Extracted features show unclassified or misclassified fringes;

Natural texture of features is lost (“flat” appearance); Depending on robustness of feature extraction, time

evolution of images is not necessarily very stable animated sequences somewhat confusing (e.g. erratically jumping classification boundaries).

IntroRGB Rev. 3 2004-07-20 15


algorithms – an example


algorithms – an example

Product PGE03/CTTH of SAFNWC/MSG software package:Cloud Top Temperature & Height

checkerboardboundary

green fringe around blue

feature

IntroRGB Rev. 3 2004-07-20 16

RGB image composites – additive colour scheme


Attribution of images of 2 or 3 channels (or channel differences/ratios) to the individual colour (RGB) beams of the display device;

RGB display devices produce colours by adding the intensities of their colour beams optical feature extraction through result of colour addition.

FAST BUT QUITE EFFICIENT SURROGATE FOR QUANTITATIVE FEATURE EXTRACTION

IntroRGB Rev. 3 2004-07-20 17



R red beam

B bl

ue b

eam

G green beam

Click Color Selector.exe

• Tool reveals individual colour intensities adding to the colours shown in the circle

• Close tool after use (also when calling it later on again)

More on RGB colours in 00_rgb_part02.ppt

IntroRGB Rev. 3 2004-07-20 18

RGB image composites – discover colour mix

RGB image composites – discover colour mix

+

+

Channel 03

Channel 02

Channel 01

Color Selector.exe

IntroRGB Rev. 3 2004-07-20 19

blue

RGB image composites – varying enhancement


observe increasing enhancement of individual RGB colour planes on the left and resulting colour shades to the right of each image couple

in 5 steps

red

green1

IntroRGB Rev. 3 2004-07-20 20

blue



red

green


in 5 steps

2

IntroRGB Rev. 3 2004-07-20 21



red blue

green


in 5 steps

3

IntroRGB Rev. 3 2004-07-20 22

blue



red

green


in 5 steps

4

IntroRGB Rev. 3 2004-07-20 23



red blue

green


in 5 steps

5

IntroRGB Rev. 3 2004-07-20 24

RGB image composites – how to do

RGB image composites – how to do

Optimum (and stable) colouring of RGB image composites depends on some manipulations:

Proper enhancement of individual colour channels requires: Some stretching of the intensity ranges; Reflectivity enhancement at lower solar angles applying

e.g. sun angle compensation or histogramme equalisation;

Selection of either P or S mode for IR channels. Attribution of images to individual colour beams

depends on: Reproduction of RGB schemes inherited from other

imagers; Permutation among colour beams of individual images

more or less pleasant / high-contrast appearance of RGB image composite.

More

on e

nhance

ment

in 0

0_r

gb_p

art

03

.ppt

IntroRGB Rev. 3 2004-07-20 25

RGB image composites – pros and cons

RGB image composites – pros and cons

Drawbacks: Much more subtle colour scheme compared to

discrete LUT used in quantitative image products interpretation more difficult;

RGBs using solar channels loose colour near dawn/dusk (even with reflectivity enhancement).

Advantages: Processes “on the fly”; Preserves “natural look” of images by retaining

original textures (in particular for clouds); Preserves spatial and temporal continuity

allowing for smooth animation of RGB image sequences.

IntroRGB Rev. 3 2004-07-20 26

RGB image composites – the classical solar case

RGB image composites – the classical solar case

Reveals fog, ice clouds and snow Channel attribution: R 03 G 02 B 01

Color Selector.exe

IntroRGB Rev. 3 2004-07-20 27

RGB image composites – more complex examplesRGB image composites – more complex examples

Reveals some cloud properties

Channel attribution:R 01 G 04 B 09

Channels 04 and 09 rendered in P mode!

Reveals atmospheric, cloud and surface features

Channel attribution:R 06-05 G 04-09 B 03-01

Color Selector.exe

IntroRGB Rev. 3 2004-07-20 28

RGB image composites – using HRV (channel 12)


In order to preserve high resolution of HRV channel assign it to 2 colour beams (using only one colour beam blurs the image too much);

Attributing it to beams R and G is preferred rendering close to natural colours for surface features;

Beam B is then free for any other SEVIRI channel properly magnified (zoom factor of 3).

Assigning an IR window channel beam B (as a temperature profile surrogate) adds height information to a detailed cloud view. Applying IR window channel in P mode renders closer to natural look when compared to S mode.

IntroRGB Rev. 3 2004-07-20 29



Reveals fine details of snow cover and low clouds / fog

Colour attribution: R 12, G 12, B 03

Reveals fine details ice (convective) clouds

Colour attribution: R 12, G 12, B 09 (09 rendered in S mode)

Color Selector.exe

IntroRGB Rev. 3 2004-07-20 30

First cut of recommended RGB image composites

First cut of recommended RGB image composites

Convection 01,03,09

01,03,10 01,04,09

01,04,10 03,04,09

03,04,10

HRV (channel) 12,12,04 12,12,09 12,12,03

Dust 01,03,04 03,02,01

Vegetation 03,02,01

Fire/Smoke 03,02,01 04,02,01

Channel differences 06-05,04-09,03-01 10-09,09-04,09 10-09,09-04,06-05

More on recommended composites and their interpretation in 00_rgb_part[04/05/06].ppt

IntroRGB Rev. 3 2004-07-20 31

Summary of RGB image composites

Summary of RGB image composites

Fast technique for feature enhancement exploiting additive colour scheme of RGB displays;

May require simple manipulation to obtain optimum colouring (choice of P or S mode for IR channels!);

More complex RGB schemes may require some time to get acquainted with;

Some RGB schemes may be inherited from other imagers (e.g. AVHRR or MODIS);

Combination of an IR channel with HRV feasible and much informative;

RGB image composites retain natural texture of single channel images;

RGB image composites remain coherent in time and space, i.e. ideal for animation of image sequences.

IntroRGB Rev. 3 2004-07-20 32

THE THE ENDENDTHE THE ENDEND

introrgb rev. 3 2004-07-20 1 introduction to rgb image composites [email protected]...

Documents

slide slide

msgseviri channels

mfg channels

images of individual

solar ir

p mode grey shades

simple grey wedge

output colour